Two admissions officers, four hundred inquiry emails, and a fee structure only the bursar fully understands.

The situation

Why the inbox drowns every season

The school's website has a fees page, but it shows last year's table for two of the five grades and says "contact admissions for details" everywhere the structure gets complicated. So parents contact admissions. The two officers answer from memory and old email threads, which means slightly different answers to the same question — and occasionally an answer that the bursar has to walk back because an installment surcharge or a scholarship condition was missed.

The fee logic itself is genuinely complex: base fees per grade, a sibling discount that applies to the second child but only on tuition, three installment plans with different surcharges, scholarship tiers that cap combined discounts, and a transport fee that depends on route zone. All of it lives in the bursar's spreadsheet and head. Nobody else can compute a family's actual number with confidence.

The school wants faster, consistent responses — but a wrong fee quote in writing is a commitment a parent will hold them to, so nothing can go out without a person confirming it.

The work

What Taz builds for this school

Taz turns the real inquiry history and the bursar's fee logic into structured, reviewable sources, then prototypes a drafting workflow where AI prepares and a human approves.

• Inquiry mining: Last season's inquiry emails are categorized to find the actual top questions and every variant of how parents phrase them — not what the school assumes gets asked, but what the archive proves gets asked.
• Canonical FAQ construction: Each recurring question gets one approved answer, written with the admissions head, with effective dates and the source of each fact. The FAQ doubles as new website content — the same thirty answers that clog the inbox become public, structured, and AI-readable.
• Fee logic structuring: The bursar's rules become an explicit, testable rule set — base fees, sibling discount conditions, installment surcharges, scholarship caps, transport zones — validated by re-computing a sample of last year's actual fee letters and reconciling every mismatch with the bursar.
• Draft-and-approve workflow: A prototype where an incoming inquiry is classified, a response is drafted from the canonical FAQ and fee rules with the computation shown, and an officer reviews, edits, and sends. The AI never sends; fee quotes always show their working.
• Escalation boundaries: Written rules for what never gets auto-drafted as routine — special-needs discussions, fee negotiations, complaints, and anything the FAQ doesn't cover with an approved answer.

The deliverable

What the school receives

• Inquiry analysis report — the real top questions ranked by volume, with phrasing variants and season timing, from the school's own archive.
• Canonical FAQ bundle — approved answers with effective dates and sources, formatted for both the response workflow and direct publication on the website.
• Structured fee rulebook — the complete fee logic as explicit rules, validated against real fee letters, readable by the bursar and executable by the workflow.
• Working prototype — classify, draft, show-the-working, human-approve — running on the school's real inquiry examples, with setup and operating instructions.
• Boundary and escalation document — what the workflow answers, what it drafts for review, and what it refuses to touch.